This invention relates to a technology for increasing reliability of communication carried out by OSes and application programs operating on logical partitions set on a computer.
As information systems become large and complicated, operation/management costs of servers constituting the information systems increase. To address this problem, there is known “server consolidation” which assigns existing physical servers to virtual servers to reduce the number of the physical servers. There is known “server virtualization” as means for the server consolidation.
The server virtualization is a technology to divide one computer resource into multiple logical partitions. An OS (i.e., guest OS) runs in the respective partitions, and the respective partitions function as virtual servers. Software for generating logical partitions, and controlling the generated partitions is a virtual machine monitor (hereinafter, referred to as “VMM”). It is possible to construct an information system by means of the guest OSes operating in the respective partitions generated by the VMM where the respective partitions function as independent servers.
In this information system, if a failure occurs to the VMM, the failure possibly exerts a crucial influence on the information system. It is thus important to monitor a failure of the VMM in order to secure the reliability of the information system. Moreover, missed detection and excessive detection of a failure pose problems in the failure monitoring, and high reliability is thus essential for the failure monitoring.
In order to increase the reliability of the failure monitoring, resources relating to the failure monitoring are to be multiplexed. The resources relating to the failure monitoring include management servers and communication paths to the management servers.
As for the management servers, the management servers should be multiplexed. On the other hand, as for the communication paths to the management servers, there poses the following problem as a result of the application program of the server virtualization.
In other words, since the server virtualization generates multiple partitions on one computer, the OSes operating on the respective partitions use I/O devices. As a result, if dedicated communication paths to the management servers are provided, and these communication paths are multiplexed, the I/O devices are consumed. As a result, the number of servers which can be consolidated by means of the server consolidation decreases, resulting in a decrease of the effect to reduce the operation/management costs.
Therefore, there arises an object to increase the reliability of the communication paths to the management servers without consuming the I/O devices.
In order to attain this object, there are two known conventional technologies which include a technology to restrain the consumption of the I/O devices in the server virtualization, and a technology to increase the reliability of the communication paths to the management servers.
According to the technology to restrain the consumption of the I/O devices in the server virtualization, there is provided a service OS (Hosting Partition), and this OS directly controls the physical I/O devices. In other partitions, when there is received an access to a virtual I/O device, the VMM notifies the service OS of the access, and an access to a physical I/O device is carried out. Consequently, the multiple partitions can share a single I/O device, and the consumption of the I/O devices can be restrained (refer to U.S. Pat. No. 6,725,284). A technology used to share a network interface card (NIC), which is an I/O device, is referred to as “NIC sharing technology”.
Moreover, as a technology to increase the reliability of the communication paths to the management servers is know multiplexing. Generally, the management servers are also multiplexed at the time of multiplexing of the communication paths in order not to stop the monitoring even if a failure occurs to each management server. Moreover, when multiplexed communication paths and management servers are used, and there is detected a failure (such as a communication error) in the management server or the communication path, a failed management server or communication path is switched to other management server or communication path. This technology is referred to as “Fail Over”.